PEIP is a technical non-starter for several reasons:
1. Not enough room in the IP header to record the path.
2. Changing the packet size in flight would greatly exacerbate the impact of the PMTUD design error in normal operations.
3. The router data plane is a poor location for any kind of complex programming.
4. The same people who have failed to implement BCP38 would have to implement the much more difficult PEIP.
5. It's whack-a-mole. The nature of the attacks is evolving from spoofed source addresses to distributed botnets with each bot performing a complete IP transaction with its own IP address. If everybody implemented BCP38 tomorrow, theses newer kinds of DDOS attacks would continue unabated.